Staged O 2 Introduction and Selective H 2 Combustion during Catalytic Reactions of Alkanes on Cation-Exchanged H-ZSM5

Transcription

1 3680 Ind. Eng. Chem. Res. 2003, 42, KINETICS, CATALYSIS, AND REACTION ENGINEERING Staged O 2 Introduction and Selective H 2 Combustion during Catalytic Reactions of Alkanes on Cation-Exchanged H-ZSM5 Toshio Waku, Sara Y. Yu, and Enrique Iglesia* Department of Chemical Engineering, University of California at Berkeley, Berkeley, California Stoichiometric reactions can be used to remove H 2 and the associated kinetic and thermodynamic barriers that lead to low dehydrocyclodimerization selectivity during alkane reactions on cationmodified H-ZSM5. O 2 co-reactants can form H 2 O in exothermic reactions that balance the enthalpy of endothermic dehydrogenation steps. O 2 reacts preferentially with H 2 via homogeneous and heterogeneous pathways, but also with hydrocarbons as H 2 is depleted; thus, it must be gradually introduced as H 2 forms in dehydrogenation reactions. Staged O 2 protocols significantly increased aromatics yields during C 3 H 8 reactions on unexchanged and on Ga- and Zn-exchanged H-ZSM5. On Ga/H-ZSM5, the maximum aromatic yields increased from 54% to 68% and aromatization/cracking selectivity ratios increased from 2.1 to 3.9 when O 2 was introduced gradually into a gradientless batch reactor as H 2 formed. O 2 introduced was converted to H 2 O with >95% selectivity; an equivalent amount of O 2 initially added with C 3 H 8 led to low H 2 O selectivities (<60%). Similar effects of O 2 addition and of staging protocols were observed for alkane reactions on H-ZSM5 and Zn/H-ZSM5. Staging strategies led to selective use of O 2 to remove thermodynamic and kinetic bottlenecks and to unprecedented aromatics yields during alkane reactions on cation-exchanged H-ZSM5. Introduction The selective conversion of light alkanes to alkenes and aromatics provides an attractive route to useful chemicals and fuels. The presence of exchanged cations, such as Ga, 1-10 Zn, and Pt, 21,22 within shapeselective channels in H-ZSM5 increases the rate and selectivity of alkane dehydrocyclodimerization reactions that form C 6 -C 9 aromatics, while decreasing the selectivity to hydrogen-rich unreactive products, such as CH 4 and C 2 H 6, which form via cracking reactions. Ga/H- ZSM5 catalysts provide the basis for the CYCLAR process. 23 Mole et al. first proposed that Zn cations act as hydrogen recombination sites in dehydrogenation steps required to form aromatics. 13 Kinetic and isotopic tracer studies later confirmed this proposal The required bifunctional pathways involve alkane and alkene reactions on zeolitic acid sites and dehydrogenation steps aided by recombinative desorption steps on exchanged Zn, Ga, Co, and other cations. The rate-limiting nature of hydrogen removal steps leads to high hydrogen virtual pressures at catalyst surfaces, which have been measured from the deuterium content in products formed from C 3 H 8 /D 2 reactants. 18 Exchanged cations increase the rate of communication between surface and gas-phase hydrogen pools by acting as recombinative * To whom correspondence should be addressed. Phone: (510) Fax: (510) iglesia@ cchem.berkeley.edu. Permanent address: Central Technical Research Laboratory, Nippon Oil Corporation, Yokohama , Japan. Current address: Argonne National Laboratories, 9700 S. Cass Ave., Argonne, IL desorption sites. In this manner, cations favor the formation of hydrogen-deficient adsorbed intermediates, which in turn favor the formation and desorption of dehydrogenation and aromatization products and also of unreactive unsaturated residues that lead to site deactivation. The hydrogen removal bottlenecks prevalent during alkane reactions on these catalysts limit alkane conversion rates, predominately by reversing the required C-H bond activation steps and by imposing an unfavorable overall equilibrium for the formation of alkene intermediates on surfaces and of aromatic products. As a result, molecules that react with adsorbed hydrogen, with hydrogen-rich intermediates, or with their gasphase precursors (e.g., H 2 ) can increase aromatics synthesis rate and selectivity. Fujimoto et al. reported that alkane/o 2 reactant mixtures led to higher aromatic selectivities than pure alkane reactants on H-ZSM5 and Ga/H-ZSM5. 5 Iglesia and Baumgartner showed that O 2, CO, and CO 2 increased the selectivity of propane conversion to aromatics on H-ZSM5 8 by acting as stoichiometric hydrogen acceptors for surface and gasphase hydrogen species. In all of these studies, O 2 reactions with H 2 lead to higher aromatics selectivity, but a large fraction of the O 2 co-reactants was consumed via unselective hydrocarbon combustion reactions to form CO and CO 2. Ga/H-ZSM5 and Zn/H-ZSM5 are the most active and selective catalysts for alkane dehydrocyclodimerization. Aromatics yields as high as 53% have been reported on Ga/H-ZSM5 3 using propane reactants and relatively high temperatures (823 K, 1 bar C 3 H 8 ). Zn/H-ZSM5 gives lower reaction rates and aromatics selectivities; the highest reported aromatics yields are below 50%, /ie030255w CCC: $ American Chemical Society Published on Web 06/27/2003

2 Ind. Eng. Chem. Res., Vol. 42, No. 16, but high aromatics selectivities (>90%) have been reported at very low but unspecified propane conversion levels. 20 These limiting yields as conversion increases reflect the selective kinetic and thermodynamic inhibition of the dehydrogenation pathways required to form aromatics and the concomitant increase in cracking selectivity as H 2 pressures increase with increasing alkane conversion. Here, we exploit the ability of semibatch reactors with continuous O 2 introduction to conveniently and accurately simulate the behavior of plug-flow tubular reactors with staged O 2 feeds in an effort to determine the kinetic and thermodynamic consequences of concurrent hydrogen combustion during propane reactions on H-ZSM5, Zn/H-ZSM5, and Ga/H-ZSM5. Controlled O 2 introduction strategies led to selective hydrogen combustion and to the removal of the kinetic and thermodynamic bottlenecks that limit aromatics yields. The availability of O 2 only as H 2 is formed is critical in our selective H 2 combustion strategies. We report oxygen staging protocols that increase aromatization to cracking selectivity ratios by a factor of 2 and lead to very selective use of O 2 co-reactants in hydrogen removal and to unprecedented aromatics yields ( 70%), even at relatively low temperatures (773 K). Experimental Methods Catalyst Synthesis and Characterization. Na- ZSM5 (Zeochem, Si/Al ) 14.5) was exchanged three times with a solution of ammonium nitrate (Fisher, Certified ACS, >98.0%) at 353 K for 16 h, then filtered, and washed with deionized water to form NH 4 -ZSM5. These NH 4 -ZSM5 samples were treated in ambient air at 398 K overnight and then in flowing dry air (Airgas, zero grade) at 773 K for 20 h to form H-ZSM5. Ga cations were introduced onto H-ZSM5 by incipient wetness impregnation using a M Ga(NO 3 ) 3 solution (Aldrich, >99.9%). The samples were then dried and treated in flowing dry air at 773 K for 20 h. Previous studies have shown that this procedure leads to Ga exchange as Ga + during subsequent reduction in H 2 at 773 K for 1-2 h. 9 Zn/H-ZSM5 samples were prepared by ion exchange of H-ZSM5 with a M solution of Zn(NO 3 ) 2 (Aldrich, >98.0%) at 343 K. The exchanged samples were filtered, washed with deionized water, dried in flowing dry air at 383 K for 20 h, and then treated in flowing dry air at 773 K for 20 h. 12 Si, Al, Na, and Ga or Zn contents were measured by atomic emission spectroscopy (Galbraith Laboratories, Inc.). H-ZSM5 samples contained traces of residual Na (0.14 wt %; Na/Al ) 0.064). Ga/H-ZSM5 contained 1.8 wt % Ga, corresponding to a Ga to framework Al atomic ratio of Zn/H-ZSM5 contained 1.0 wt % Zn and a Zn/Al F atomic ratio of The number of residual OH groups, measured by isotopic exchange with D 2, were similar in Zn and Ga samples because of the respective exchange stoichiometries of isolated Ga + and Zn 2+ cations. Catalytic Reaction Rates and Selectivities. All reaction rate and selectivity measurements were carried out using a gradientless recirculating batch reactor. 7 The reactor contents (520 cm 3 ) were circulated at >2 cm 3 s -1 to ensure differential reaction conditions (<1% reactant conversion per pass). Catalyst samples ( g) were treated at 773 K in flowing dry air (Airgas, zero grade) for 1 h before catalytic experiments. Ga/H- ZSM5 samples were also treated in flowing H 2 at 773 K for 2 h; this treatment led to the reduction of unexchanged Ga 2 O 3 to form volatile Ga 2 O species, which migrate into ZSM5 channels and replace some of the acidic H + species in H-ZSM5. 6,10 Propane reaction rates and selectivities were measured at 773 K using propane (20.0 kpa, Praxair, %) and three reactant introduction protocols: (i) C 3 H 8 reactants (20.0 kpa). (ii) C 3 H 8 (20.0 kpa) and O 2 (13.4 kpa) (Scott Specialty Gases, Inc., %) concurrently introduced at the start of the experiment (co-feed mode). (iii) C 3 H 8 (20.0 kpa) introduced at the start and O 2 added gradually at a constant rate ( mmol/ g-atom of Al s) (staged feed mode). Oxygen staging intensity is defined as the ratio of the total amount of introduced O 2 to that required to consume all of the H 2 formed during alkane reactions. A staging intensity of 1.0 is defined as the stoichiometric case. Ar (1 kpa; research grade > %) was used as an internal standard and the reactor was initially kept at ambient pressure using He (balance; Airgas, research grade > %) as an inert diluent. O 2 was introduced via a quartz capillary column (60-m length, 250- µm diameter) at a rate controlled by the pressure imposed by an upstream regulator. Reactant and product concentrations were measured using on-line gas chromatography (Hewlett-Packard, Model 5890). Samples were collected using a ten-port valve with two sample loops residing within the recirculation path. The samples were injected into capillary (HP-1 methyl-silicone column, 50 m, 320 µm) and packed (Supelco, 1000 Carboxen) columns. Flame ionization was used to detect hydrocarbons eluting from the capillary column. The packed column effluent was analyzed by thermal conductivity to measure Ar, N 2, O 2, CO, and CO 2. Batch reactor data are shown as product site yields (mol/g-atom of Al). Reaction rates were calculated from the slopes of these data and reported as molar propane conversion rates (total or leading to a given product per g-atom of Al). Rates are calculated from the slopes of the line connecting each point to the origin; these rates correspond to those measured in a mathematically equivalent plug-flow reactor after that residence time. The mathematical analogy between temporal staging of O 2 in a batch reactor and spatial O 2 staging in a plugflow reactor is described in detail elsewhere. 24 Hydrocarbon selectivities are reported on a carbon basis as the percentage of the converted propane appearing as a given reaction product. Hydrogen selectivities are reported as the percentage of the hydrogen in the converted propane appearing within each product. Selectivities are also given on an oxygen basis for O 2 introduction experiments; they are defined as the fraction of the O 2 introduced that appears as H 2 OorCO x (CO and CO 2 ). These oxygen-based selectivities represent a measure of the relative effectiveness with which O 2 is used to combust hydrogen and hydrocarbons. The approach to equilibrium (η) for each product formation reaction was determined from the thermodynamics of the corresponding stoichiometric reaction, as illustrated below for benzene: C 3 H 8 f 1 / 2 C 6 H / 2 H 2 [P C6 H 6 ] 1/2 [P H2 ] 5/2 η ) [P C3 H 8 ] 1 K EQ

3 3682 Ind. Eng. Chem. Res., Vol. 42, No. 16, 2003 Table 1. Effects of Oxygen Staging on Propane Dehydrocyclodimerization Turnover Rates and Product Distribution in Gradientless Batch Reactors [773 K, 20 kpa Propane, Balance He] 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25) 1.0 wt % Zn/H-ZSM5 (Zn/Al ) 0.14) H-ZSM5 (Si/Al ) 14.5) oxygen introduction rate (mmol of O 2/g-atom of Al s) oxygen staging intensity propane conversion (%) propane to hydrocarbons turnover rate (per Al, 10-3 s -1 ) propane to aromatics turnover rate (per Al, 10-3 s -1 ) propane to carbon oxides turnover rate (per Al, 10-3 s -1 ) carbon selectivity (%) methane ethene ethane propene C 4-C C 6+ aromatics benzene toluene C 8 aromatics C 9+ aromatics hydrogen selectivity (%) H oxygen selectivity (%) H 2O CO CO Figure 1. Effects of cations on C 6+ aromatics selectivity during propane dehydrocyclodimerization in gradientless batch reactors [773 K, 20 kpa propane, balance He, 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25); 1.0 wt % Zn/H-ZSM5 (Zn/Al ) 0.14); H-ZSM5 (Si/Al ) 14.5)]. K EQ are equilibrium constants obtained from literature thermodynamic data. 25 The value of η is zero far away from equilibrium and approaches unity as equilibrium is reached; thus, (1 - η) represents the fractional distance from equilibrium for a given reaction, which increases with increasing reaction time. Results and Discussion Propane Dehydrocyclodimerization on H-ZSM5, Zn/H-ZSM5, and Ga/H-ZSM5. Ga and Zn cations increased propane turnover rates and the rate and selectivity for aromatics formation (Table 1). These effects are illustrated at two propane conversion levels in Table 1; they reflect the contribution of exchanged cations to hydrogen removal as H ,17,18 The fraction of the hydrogen in the converted propane that appears as H 2 increased significantly as some of the protons in Figure 2. Effects of temperature on C 6+ aromatics selectivity during propane dehydrocyclodimerization in gradientless batch reactors [20 kpa propane, balance He, 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25)]. H-ZSM5 were replaced by Ga or Zn cations (Table 1). Reaction selectivities on different catalysts must be compared at similar propane conversions because the sequential nature of dehydrocyclodimerization pathways, the intermediate role of propene, and the approach to equilibrium for some reactions as H 2 forms combine to lead to marked changes in selectivity as conversion increases (Figure 1, Table 1). At all conversion levels, Ga/H-ZSM5 shows the highest C 6+ aromatics selectivity; these selectivities, however, reach a maximum value of 67% and then decrease as propane conversions increase above 60% (Figure 2). These maximum attainable selectivities reflect the stronger kinetic and thermodynamic effects of H 2 concentrations on propane to aromatics reaction rates compared with their effects on the rates of parallel cracking steps. Similar trends were observed at lower temperatures

4 Ind. Eng. Chem. Res., Vol. 42, No. 16, Table 2. Approaches to Equilibrium during Propane Dehydrocyclodimerization Reactions in Gradientless Batch Reactors [773 K, 20 kpa Propane, Balance He, 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25); 1.0 wt % Zn/H-ZSM5 (Zn/Al ) 0.14); H-ZSM5 (Si/Al ) 14.5)] Ga/H-ZSM5 Zn/H-ZSM5 H-ZSM5 propane conversion (%) η (C 3H 8 f 1 / 2C 6H / 2H 2) η (C 3H 8 f C 3H 6 + H 2) η (C 3H 6 f 1 / 2C 6H / 2H 2) η (C 3H 6 f 3 / 7C 7H / 7H 2) η (C 2H 4 f 1 / 3C 6H 6 + H 2) η (C 2H 4 f 2 / 7C 7H / 7H 2) (673 K; Figure 2), but maximum selectivities are lower ( 55%) and occur at lower conversions (30%). An indication of the extent to which various reactions are limited by equilibrium can be obtained from the approach to equilibrium data shown in Table 2. Propane conversion to benzene is far from equilibrium at all conversions, suggesting that net benzene formation rates are largely unaffected by its reverse reaction. Propene to benzene and propane to toluene conversions are also far from equilibrium. Other reactions, such as ethene conversion to benzene, are near equilibrium at high propane conversions (>60%) on Ga/H-ZSM5. The approaches to equilibrium for ethene to toluene reactions are smaller than those for ethene conversion to benzene (0.26 vs 1.0, at 80.8% conversion; Ga/H-ZSM5), but reverse reactions cannot be neglected, even for the former reaction. Thus, some ethene formed in primary propane cracking steps cannot convert to aromatics because of thermodynamic constraints; instead, they hydrogenate to form unreactive ethane molecules as H 2 pressures increase with increasing conversion. 26 The net rate for all reactions with nonzero η values will increase as H 2 is removed in combustion reactions favored by thermodynamics. The presence of H 2 during propane reactions leads to a shift in selectivity toward hydrogen-rich cracking products and to a decrease in dehydrogenation and aromatization rates on Zn/H-ZSM5 18 and Co/H-ZSM5. 26 These effects of H 2 are much stronger on cationexchanged H-ZSM5 than on H-ZSM5 because cations activate H 2 and introduce hydrogen into the surface pool of reactive intermediates. At low propane conversions, Ga and Zn remove hydrogen from this pool as H 2, but as alkane conversion and H 2 concentrations increase, the reverse of this reaction leads to dissociative adsorption and to unfavorable thermodynamics for required surface reactions. In effect, as one or more of the required dehydrogenation steps approach equilibrium, H 2 leads to lower net reaction rates because the rate of the corresponding reverse reactions gradually increases. In contrast, the rate of cracking reactions that form CH 4 and of C 2 H 4 hydrogenation to form C 2 H 6 are either unaffected by H 2 or increase with increasing H 2 pressure because these reactions remain far from equilibrium at all propane conversion levels. Effects of Hydrogen Removal by Staged O 2 Introduction on C 3 H 8 Reactions. Selective stoichiometric reactions of H 2 during propane reactions can remove these kinetic and thermodynamic barriers. The reported beneficial effects of O 2 on aromatics selectivity combined with the observed undesired hydrocarbon combustion reactions 5,8 led us to explore the staging of O 2 co-reactants so as to match the rate at which H 2 becomes available from C 3 H 8 reactions. The effects of O 2 staging were examined using a gradientless batch reactor that rigorously simulates the gradual injection Figure 3. Effects of oxygen staging on product selectivities on 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25) in gradientless batch reactors [773 K, 20 kpa propane, 3.3 mmol/g-atom of Al s O 2, balance He]. of O 2 via multiple injectors or oxygen-permeable walls in plug-flow reactors. O 2 introduction rates were initially chosen based on average H 2 formation rates (staging intensity of unity) and then varied around these stoichiometric values. At low conversions (<60%), staged O 2 introduction did not significantly influence propane conversion rates and selectivities on 1.8 wt % Ga/H-ZSM5 (Figure 3), but led to higher rates and selectivities at higher C 3 H 8 conversions. The highest achievable C 6+ aromatics selectivity was significantly higher ( 80%) and was reached at higher conversions (85%) than with pure C 3 H 8 reactants (67% at 70% conversions). The maximum attainable yields of C 6+ aromatics increased from 54% for C 3 H 8 reactants to 68% with stoichiometric O 2 staging on Ga/ H-ZSM5. Such C 6+ aromatics yields are much higher than those previously reported for propane reactions on Ga/H-ZSM5 or Zn/H-ZSM5. 3,19 Cracking selectivities were significantly lower with staged O 2 injection than with pure C 3 H 8 reactants, especially at high conversions (Figure 3). C 3 H 6 and C 4 -C 5 selectivities were less affected by O 2 addition and H 2 removal than cracking selectivities (Table 1). O 2 co-reactants decreased C 3 H 6 (3.5% to 2.9%) and C 4 -C 5 (1.4% to 0.8%) selectivities at 60% conversion on Ga/H-ZSM5. C 3 H 6 and C 4 -C 5 products ultimately convert to C 6+ aromatics; thus, a rigorous measure of reactant efficiency is given by the ratio of desired aromatic products to undesired cracking products, especially unreactive ethane and methane. Figure 4 shows that O 2 addition strategies lead to almost a 2-fold increase in this ratio at the high C 3 H 8 conversion levels required for practical C 6+ aromatic yields. Propane reaction pathways include primary dehydrogenation steps that form C 3 H 6 and primary cracking steps that form CH 4 and C 2 H 4 directly from propane, as shown in Scheme Secondary products include C 2 H 6 and C 6+ aromatics. 26 O 2 addition did not strongly

5 3684 Ind. Eng. Chem. Res., Vol. 42, No. 16, 2003 Figure 4. C 6+ aromatics to C 1 + C 2 selectivity ratio versus propane to hydrocarbons conversions [773 K, 20 kpa Propane, 13 kpa O 2 (co-feed), 3.3, 6.6 mmol/g-atom of Al s O 2, balance He, 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25)]. Figure 6. (a) Methane and (b) C 2 (the sum of ethene and ethane) site yields during propane dehydrocyclodimerization on 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25) [773 K, 20 kpa propane, 3.3 mmol/gatom of Al s O 2, balance He]. Table 3. Effects of Oxygen Staging on the Ratios of C 2 to C 1 Site Yields on 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25) in Gradientless Batch Reactors [773 K, 20 kpa Propane, 3.3 mmol/g-atom of Al s O 2, Balance He] C 3H 8 C 3H 8/staged O 2 propane conversion % C 2/C 1 site yields ratio propane conversion % C 2/C 1 site yields ratio Scheme 1. Reaction Pathways for Propane Conversion on ZSM5-Based Catalysts. Adapted from Ref 26 Figure 5. Propane to (a) hydrocarbons turnovers and (b) carbon oxides turnovers during propane dehydrocyclodimerization on 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25) [773 K, 20 kpa propane, 13 kpa O 2 (co-feed), 3.3, 6.6 mmol/g-atom of Al s O 2, balance He]. influence C 3 H 8 conversion rates (Figure 5a), consistent with the fact that primary reactions are far from equilibrium (Table 2) and thus unaffected by H 2 removal. O 2 co-reactants decreased C 2 site yields (Figure 6b), especially as conversion increased with increasing contact time, but did not affect CH 4 formation rates (Figure 6a). In all cases, the C 2 /C 1 molar ratio was smaller than 2, the value expected from primary cracking steps (Table 3, Scheme 1). Thus, primary cracking rates were not affected by H 2 concentration or by its removal with O 2, but the latter influenced the relative rates of C 2 H 4 hydrogenation and aromatization reactions. Indeed, the near-equilibrated nature of ethene conversion to aromatics prevents the net incorporation of C 2 H 4 into the aromatics pool, while the high prevalent H 2 concentrations with pure C 3 H 8 reactants favors its hydrogenation to form unreactive C 2 H 6. Thus, O 2 addition shifts the equilibrium for ethene conversion to aromatics, while concurrently removing the H 2 coreactants required for parallel undesired hydrogenation steps. The efficiency of O 2 co-reactants in removing H 2 is given by the percentage of the converted O 2 that forms

6 Ind. Eng. Chem. Res., Vol. 42, No. 16, Figure 7. (a) Total amounts of introduced oxygen and (b) residual oxygen during propane dehydrocyclodimerization on 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25) [773 K, 20 kpa propane, 13 kpa O 2 (co-feed), 3.3, 6.6 mmol/g-atom of Al s O 2, balance He]. Figure 8. H 2O selectivity (%O base) during propane dehydrocyclodimerization on 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25) in recirculating batch reactors [773 K, 20 kpa propane, 13 kpa O 2 (co-feed), 3.3, 6.6 mmol/g-atom of Al s O 2, balance He]. H 2 O instead of CO x (CO and CO 2 ). Figure 7 shows the added and residual O 2 amounts for a staging intensity of unity on Ga/H-ZSM5. Residual O 2 concentrations are very low (<0.11 µmol/cm 3 ) throughout the entire range of conversion and contact time. Figure 8 shows that H 2 O selectivities for this staging intensity are 95-98%, indicating that H 2 removal by O 2 is remarkably selective, as long as H 2 is present to deplete O 2 co-reactants. H 2 O selectivities decreased slightly with increasing conversion, apparently because reaction products are more reactive than C 3 H 8 in combustion reactions and these products may also inhibit homogeneous H 2 combustion rates via radical trapping. 27 In addition, the decrease in H 2 formation rates with increasing C 3 H 8 conversion leads to the ultimate accumulation of some unreacted O 2 and to its greater involvement in hydrocarbon oxidation. Staged O 2 addition strategies lead to similar selectivity and yield improvements on Zn/H-ZSM5 and H-ZSM5 (Table 1) and these results are discussed in a later section. Effects of O 2 Staging and Co-feed Strategies on Ga/H-ZSM5. In view of the marked improvements in hydrocarbon and O 2 selectivities achieved by stoichiometric staging of O 2 co-reactants, we examined the effects of staging intensity on reaction rates and selectivity. Experiments were carried out using 20 kpa C 3 H 8 by either introducing O 2 initially with propane reactants (13.4 kpa) or using an O 2 staging intensity of 1.0 (3.3 mmol/g-atom of Al s). These data were compared with those obtained in co-feed mode, in which the amount of O 2 added over 20 ks during staged experiments was added initially together with C 3 H 8 reactants. Residual O 2 concentrations are initially very high in this co-feed mode, but they decreased rapidly as O 2 is predominately consumed in fast hydrocarbon combustion reactions (Figure 7b). O 2 concentrations become undetectable at contact times required for 35% propane conversion. In the absence of significant H 2 concentration, reactions of O 2 are fast and unselective and O 2 remains available only for contact times leading to relatively low C 3 H 8 conversions, for which H 2 removal has minor kinetic and thermodynamic consequences. O 2 is available continuously and present at much lower average concentrations when it is introduced continuously as propane reacts and H 2 is formed. The depletion of O 2 at 35% C 3 H 8 conversions causes subsequent propane reactions to occur without the benefit of H 2 removal; yet, it is precisely at higher conversions that rates and selectivities would benefit from selective H 2 reactions with O 2 (Figure 4). At low conversions, co-feed and staging O 2 strategies lead to similar effects on aromatic/cracking selectivity ratios, but these effects disappear as O 2 is depleted via unselective reactions in the co-feed mode, while they continue at higher conversion because of the continuous availability of O 2 during staged O 2 operation (Figure 4). The high initial O 2 concentrations prevalent in cofeed operating modes lead to the unselective use of O 2 in combustion of propane and of its reaction products (Figures 5b and 8). Figure 5b shows CO x formation turnovers for co-feed and stoichiometric staging cases. CO x formation rates are initially very high in co-feed mode, but decrease with contact time as O 2 is depleted. CO x formation rates are much lower at the lower O 2 concentrations prevalent during staged O 2 introduction. More importantly, the selectivity for O 2 consumption in H 2 O formation is very low in the co-feed mode (Figure 8) because O 2 concentrations are highest during the early reaction stages, when H 2 is unavailable (Figure 7b). O 2 concentrations decrease rapidly to undetectable levels as H 2 becomes available with increasing C 3 H 8 conversion. O 2 is used selectively to form H 2 O during stoichiometric O 2 staging (Figure 8) and H 2 O selectivities remain above 95% at all conversion levels. In contrast, initial H 2 O selectivities are very low ( 57%) in the co-feed mode and increase to 70% as O 2 is depleted with increasing contact time (Figure 8). Thus, staging strategies provide an attractive route for the

7 3686 Ind. Eng. Chem. Res., Vol. 42, No. 16, 2003 Table 4. Propane to Carbon Oxides Turnover Rates on 1.8% Ga/H-ZSM5 and with Empty Reactor [Catalytic Reactions: 773 K, 20 kpa Propane, 2.2 mmol/g-atom of AlsO 2, Balance He, 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25), Homogeneous Reactions: 773 K, Reactants and Products after the Catalytic Reaction for 7.2 ks, at the Same O 2 Rate as with Catalyst, Empty Reactor] Ga/H-ZSM5 homogeneous contact time (ks) carbon oxides formation rate (mol/cm 3 (heated volume), 10-9 s -1 ) H 2 combustion rate (mol/cm 3 (heated volume), 10-7 s -1 ) residual H 2 concentration (mol/cm 3 (reactor volume), 10-6 ) residual O 2 concentration (mol/cm 3 (reactor volume), 10-8 ) selective use of O 2 in propane dehydrocyclodimerization reactions and lead to much higher aromatics selectivities and O 2 reactant efficiencies than co-feed operating modes. To determine the extent to which homogeneous pathways contribute to CO x formation, reaction products and unreacted C 3 H 8 were isolated after reaction (to 55% conversion) on Ga/H-ZSM5 for 7.2 ks at 773 K using an O 2 staging intensity of 0.67 (2.2 mmol/g-atom of Al s). This mixture was then reacted in an empty reactor (2.0- cm 3 heated volume) identical to that used in catalytic experiments, while O 2 introduction was continued at the same molar rate. The results of this experiment are shown in Table 4. CO x formation and H 2 combustion rates were initially unchanged by the removal of Ga/ H-ZSM5 catalyst, indicating that homogeneous oxidation pathways are sufficiently rapid to account for all the CO x formed and all of H 2 combusted during O 2 staging. As H 2 was depleted by continuous O 2 addition, CO x formation rates increased markedly because of the resulting increase in O 2 concentrations. We conclude that tubular packed bed reactors, with void to catalyst volume ratios of 1 are likely to give even lower CO x selectivities than measured in our reactors, for which this volume ratio is 20. Effect of O 2 Introduction Rates on Propane Reaction on Ga/H-ZSM5. We have also examined the effects of O 2 staging intensity (0, 0.67, 1.0, 1.5, and 2.0) on hydrocarbon and oxygen selectivities for propane reactions on Ga/H-ZSM5. At an O 2 staging intensity of 2.0, O 2 concentrations increased gradually with time (Figure 7b) because H 2 formed in dehydrogenation reactions is quickly depleted (Figure 9) and the residual O 2 reacts only in slower hydrocarbon oxidation reactions. Staging intensities above stoichiometric values lead only to small additional improvements in aromatics/cracking selectivity ratios over those achieved with stoichiometric staging (Figure 4); they lead, however, to lower O 2 reactant efficiencies and H 2 O selectivities (Figure 8). CO x turnover rates increased markedly as staging intensities reached values above 1 (Figure 5b). The stoichiometric case provides the amount of O 2 required for H 2 depletion and avoids the presence of residual O 2 for slower competitive reactions with hydrocarbons. Aromatics/cracking selectivity ratios and H 2 O selectivities are shown at 40% propane conversion as a function of O 2 staging intensity in Figure 10. O 2 staging intensity strongly influenced selectivities for values less than 1, but only weakly at higher staging intensities. Aromatics selectivities increased monotonically with staging intensity, while H 2 O selectivities decreased sharply above stoichiometric values. Figure 9. Effects of O 2 staging intensity on hydrogen concentration during propane dehydrocyclodimerization on 1.8 wt % Ga/ H-ZSM5 (Ga/Al ) 0.25) [773 K, 20 kpa propane, 3.3, 6.6 mmol/ g-atom of Al s O 2, balance He]. Figure 10. Effects of O 2 staging intensity on C 6+ aromatics to C 1 + C 2 selectivity ratio and H 2O selectivity (%O base) on 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25) in a recirculating batch reactor [773 K, 20 kpa propane, mmol/g-atom of Al s O 2, balance He, 39-42% propane conversion]. Interestingly, staging intensities above 1 also lead to lower propane dehydrocyclodimerization rates on Ga/ H-ZSM5 (Figure 5a). This loss of activity may reflect (1) oxidation of active sites by the excess O 2, (2) loss of zeolite crystallinity or accessibility caused by higher H 2 O concentrations, or (3) deposition of unreactive organic residues at the low prevalent H 2 concentrations. Indeed, exchanged Ga + species may oxidize to GaO +, 10 but Zn/H-ZSM5, in which Zn cations remain divalent throughout treatments in H 2, propane, and O 2, showed similar losses in activity at high O 2 introduction rates. Thus, we conclude that oxidation is not responsible for the observed loss of catalytic activity. Treatment in flowing dry air at 773 K for 2 h restored 95% of the initial propane conversion rates, suggesting that loss of crystallinity is not responsible for the observed deactivation. Unsaturated organic residues are favored by the low H 2 levels prevalent at high O 2 introduction rates, especially when these low H 2 levels coexist with significant concentrations of unsaturated molecules formed in propane reactions. 18 It is expected that such residues would be favored at conditions that also lead to high aromatics selectivity, as these residues form from unsaturated adsorbed intermediates. These conclusions are consistent with the observed increase in the size of aromatic products with increasing staging intensity

8 Ind. Eng. Chem. Res., Vol. 42, No. 16, Figure 11. Effects of O 2 staging intensity on C 9+/C 6 - C 8 aromatics selectivity ratio on 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25) in recirculating batch reactors [773 K, 20 kpa propane, 3.3 (stoichiometric staging), 6.6 (excess staging) mmol/g-atom of Al s O 2, balance He]. Figure 12. Effects of oxygen staging on (a) C 6+ aromatics to C 1 + C 2 selectivity ratios; (b) propane to hydrocarbons turnovers; (c) H 2O selectivities (%O base): Without oxygen (open marks), with staged oxygen (closed marks) [773 K, 20 kpa propane, 3.3 (Ga), 1.7 (Zn), 0.75 (H) mmol/g-atom of Al s O 2, balance He, 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25); 1.0 wt % Zn/H-ZSM5 (Zn/Al ) 0.14); H-ZSM5 (Si/Al ) 14.5)]. (Figure 11). In addition, the presence of 13 CO in 12 C 3 H 8 reactants led to catalytic deactivation and to the formation of labeled carbon deposits; 8 thus, the higher CO x concentrations present at above stoichiometric staging intensities may also lead to deactivating deposits. We conclude that the benefits of selective H 2 removal with O 2 depend on some critical variables: the time (or reactor position) and the rate at which O 2 is introduced, the presence of O 2 only as H 2 becomes available from dehydrogenation reaction, and the retention of some H 2 to moderate the tendency of active sites to form large and highly unsaturated products and adsorbed intermediates. A balance among these variables is achieved when the intensity and location (or time) of O 2 introduction is controlled so as to maintain a low but nonzero concentration of H 2 throughout the reactor. This balance is typically achieved at staging intensities slightly below stoichiometric and when introduction rates can be adjusted to reflect changes in propane dehydrocyclodimerization rates as the kinetic driving force decreases along the reactor (in a tubular flow reactor) or with increasing contact time (in a semibatch reactor). Effects of O 2 Staging on Propane Dehydrocyclodimerization Reactions Catalyzied by H-ZSM5 and Zn/H-ZSM5. Here, we examine the effects of stoichiometric O 2 staging during propane reactions on H-ZSM5 and 1.0% Zn/H-ZSM5 and compare them with those described above for 1.8% Ga/H-ZSM5. Some of the relevant results are shown in Figure 12a-c and in Table 1. In each case, the staging intensity is defined as the ratio of the total amount of introduced O 2 to that required to consume all H 2 formed during alkane reactions. O 2 staging led to a significant increase in aromatics/cracking selectivity ratios on all three catalysts (Figure 12a). These effects are stronger when cations are present in H-ZSM5 because dehydrogenation reactions, and specifically ethene conversion to aromatics, a reaction that preserves ethene within the reactive alkene pool, are much closer to equilibrium than on H-ZSM5 (Table 2). The principal role of H 2, and the main benefit of its removal via selective combustion, is to shift the equilibrium of this reaction toward the formation of aromatics and the reaction selectivity away from the formation of less reactive ethane (Scheme 1). At these staging intensities, propane conversion rates are essentially unaffected by O 2 introduction (Figure 12b), suggesting that only the relative rates of secondary reactions, and not primary dehydrogenation and cracking reactions of propane, depend on H 2 concentration. On all three catalysts, benzene and C 9+ aromatics selectivities were significantly higher with O 2 coreactants than with pure C 3 H 8 (Table 1), while selectivities to toluene and xylenes were less sensitive to O 2 staging. This appears to reflect the closer approach to equilibrium for ethene conversion to benzene relative to those for ethene conversion to toluene and xylenes (Table 2). Figure 12 shows selectivities for O 2 conversion to H 2 O at different staging intensities on each catalyst. At similar O 2 staging intensities (1.3), H 2 O selectivities are

9 3688 Ind. Eng. Chem. Res., Vol. 42, No. 16, 2003 Table 5. Relative Combustion Rates of Hydrocarbons during Homogeneous Reaction [Homogeneous Reaction: 773 K, Reactants and Products after the Catalytic Reaction for 7.2 ks, at the Same O 2 Rate as with Catalyst, Empty Reactor; Catalytic Reaction: 773 K, 20 kpa Propane, 2.2 mmol/g-atom of Al s O 2, Balance He, 1.8 wt % Ga/H-ZSM5 (Ga/Al ) 0.25)] relative combustion rates CH 4 C 2H 4 C 2H 6 C 3H 6 C 3H 8 aromatics C similar on Zn/H-ZSM5 (89%) and Ga/H-ZSM5 (91%). At higher values of staging intensity (1.6), H 2 O selectivities were lower on H-ZSM5 (79%) than on Ga/H-ZSM5 (87%). Propene selectivities are higher and H 2 selectivities lower on H-ZSM5 than on Ga or Zn/H-ZSM5 (Table 1) at all conversion levels. The homogeneous oxidation first-order rate constant for propene is much higher (12.6 relative to propane) than those for propane (1.0) and aromatics (1.8). These values were obtained from disappearance rates for each hydrocarbon during homogeneous reaction experiments (Table 5). The more reactive nature of the products formed and their relatively high abundance compared with H 2 appear to account for the lower H 2 O selectivities measured on H-ZSM5 relative to those on cation-exchanged H-ZSM5. Conclusions Semibatch reactors with continuous O 2 introduction were used to probe the effects of selective H 2 combustion during propane dehydrocyclodimerization reactions on H-ZSM5, Zn/H-ZSM5, and Ga/H-ZSM5. These experiments rigorously simulate the gradual injection of O 2 along a plug-flow reactor via mutiple injectors or oxygen-permeable walls. Controlled O 2 introduction strategies led to selective combustion of the H 2 formed during propane reactions. This approach overcomes the kinetic and thermodynamic inhibition effects of H 2, which limit aromatic yields in these reactions. Specifically, selective H 2 removal increases the net rate of the nearly equilibrated conversion of the ethene formed in cracking reactions into aromatics. It also decreases the rate at which ethene converts to unreactive ethane. These O 2 staging protocols provide an attractive route for the selective use of O 2 for H 2 combustion (H 2 O selectivity >95%), which appear to proceed via homogeneous pathways. Staging intensities near unity (stoichiometric addition case) led to a 2-fold increase in the aromatization to cracking selectivity ratio and to unprecedented aromatics yields ( 70%) on Ga/H-ZSM5, even at relatively low temperature (773 K). The introduction of an equivalent amount of O 2 with propane reactants led to the unselective depletion of O 2 in hydrocarbon oxidation reactions and to the substantial absence of O 2 as H 2 is formed in propane dehydrocyclodimerization reactions. These benefits of O 2 introduction were also detected on Zn/H-ZSM5 and H-ZSM5 catalysts. On H-ZSM5, the higher selectivity to propene products, which are much more reactive than propane or aromatics in homogeneous oxidation reactions, and the lower H 2 selectivity led to less selective use of added O 2 in H 2 removal reactions. Acknowledgment One of the authors (T.W.) acknowledges the financial support of Japan Cooperation Center, Petroleum and Nippon Oil Corporation for his research activities at the University of California at Berkeley. Literature Cited (1) Gregory, R.; Kolombos, A. J. Chemical process making aromatic hydrocarbons over gallium catalyst. U.S. Patent 4,056,575, (2) Bulford, S. N.; Davies, E. E. Process for aromatizing C 3-C 8 hydrocarbon feedstocks using a gallium containing catalyst supported on certain silicas. U.S. Patent 4,157,356, (3) Kitagawa, H.; Sendoda, Y.; Ono, Y. Transformation of propane into aromatic-hydrocarbons over ZSM-5 zeolites. J. Catal. 1986, 101, 12. (4) Gnep, N. S.; Doyemet, J. Y.; Guisnet, M. Role of gallium species on the dehydrocyclodimerization of propane on ZSM5 catalysts. J. Mol. Catal. 1988, 45, 281. (5) Fujimoto, K.; Nakamura, I.; Yokota, K. Oxidative dehydrogenation of lower paraffins with ZSM-5 zeolite catalysts. Zeolites 1989, 9, 120. (6) Price, G. L.; Kanazirev, V. Ga 2O 3/HZSM-5 propane aromatization catalysts-formation of active-centers via solid-state reaction. J. Catal. 1990, 126, 267. (7) Iglesia, E.; Baumgartner, J. E.; Price, G. L. Kinetic coupling and hydrogen surface fugacities in heterogeneous catalysis. 1. Alkane reactions on Te/NaX, H-ZSM5, and Ga/H-ZSM5. J. Catal. 1992, 134, 549. (8) Iglesia, E.; Baumgartner, J. E. Hydrogen-transfer and activation of propane and methane on ZSM5-based catalysts. Catal. Lett. 1993, 21, 55. (9) Meitzner, G. D.; Iglesia, E.; Baumgartner, J. E.; Huang, E. S. The chemical-state of gallium in working alkane dehydrocyclodimerization catalysts - in situ gallium K-edge X-ray Absorption-Spectroscopy. J. Catal. 1993, 140, 209. (10) Biscardi, J. A.; Iglesia, E. Structure and function of metal cations in light alkane reactions catalyzed by modified H-ZSM5. Catal. Today 1996, 31, 207. (11) Biscardi, J. A.; Meitzner, G. D.; Iglesia, E. Structure and density of active Zn species in Zn/H-ZSM5 propane aromatization. J. Catal. 1998, 179, 192. (12) Biscardi, J. A.; Iglesia, E. Reaction pathways and ratedetermining steps in reactions of alkanes on H-ZSM5 and Zn/H- ZSM5 catalysts. J. Catal. 1999, 182, 117. (13) Mole, T.; Anderson, J. R.; Creer, G. The reaction of propane over ZSM-5-H and ZSM-5-Zn zeolite catalysts. Appl. Catal. 1985, 17, 141. (14) Ono, Y. Transformation of lower alkanes into aromatichydrocarbons over ZSM-5 zeolites. Catal. Rev.-Sci. Eng. 1992, 34 (3), 179. (15) Berndt, H.; Lietz, G.; Lucke, B.; Volter, J. Zinc promoted H-ZSM-5 catalysts for conversion of propane to aromatics. 1. Acidity and activity. Appl. Catal. A 1996, 146, 351. (16) Viswanadham, N.; Pradhan, A. R.; Ray, N.; Vishnoi, S. C.; Shanker, U.; Prasada Rao, T. S. R. Reaction pathways for the aromatization of paraffins in the presence of H-ZSM-5 and Zn/H- ZSM-5. Appl. Catal. A 1996, 137, 225. (17) Biscardi, J. A.; Iglesia, E. Non-Oxidative Reactions of Alkanes on Zn/Na-ZSM5: Structural and Catalytic Properties. Phys. Chem. Chem. Phys. 1999, 1, (18) Yu, S. Y.; Biscardi, J. A.; Iglesia, E. Kinetic relevance of hydrogen desorption steps and virtual pressures on catalytic surfaces during reactions of light alkanes. J. Phys. Chem. B 2002, 106, (19) Nicolaides, C. P.; Sincadu, N. P.; Scurrell, M. S. NAS (novel aluminosilicates) as catalysts for the aromatisation of propanes Studies of zinc and gallium modified zeolite-based systems having various extents of XRD Crystallin. Catal. Today 2002, 71, 429. (20) Lubango, L. M.; Scurrell, M. S. Light alkanes aromatization to BTX over Zn-ZSM-5 catalystssenhancements in BTX selectivity by means of a second transition metal ion. Appl. Catal. A 2002, 235, 265. (21) Inui, T.; Oakazumi, F. Propane conversion to aromatichydrocarbons on Pt/H-ZSN-5 catalysts. J. Catal. 1984, 90, 366. (22) Kwak, B. S.; Sachtler, W. M. H.; Haag, W. O. Catalytic conversion of propane to aromaticsseffects of adding Ga and/or Pt to HZSM-5. J. Catal. 1994, 149, 465. (23) Mowry, J. R.; Anderson, R. F.; Johnson, J. A. Process makes aromatics from LPG. Oil Gas J. 1985, 83, 128.

10 Ind. Eng. Chem. Res., Vol. 42, No. 16, (24) Waku, T.; Biscardi, J. A.; Iglesia, E. Catalytic dehydrogenation of alkanes on Pt/Na-[Fe]ZSM5 and staged O 2 introduction for selective H 2 removal. Submitted for publication in J. Catal. (25) Stull, D. R.; Westrum, E. F., Jr.; Sinke, G. C. The Chemical Thermodynamics of Organic Compounds; John Willey & Sons: New York, (26) Yu, S. Y.; Yu, G. J.; Li, W.; Iglesia, E. Kinetics and reaction pathways for propane dehydrogenation and aromatization on Co/ H-ZSM5 and H-ZSM5. J. Phys. Chem. B 2002, 106, (27) Hucknall, D. J. Chemistry of Hydrocarbon Combustion; Chapman and Hall: New York, Received for review March 26, 2003 Revised manuscript received May 15, 2003 Accepted May 16, 2003 IE030255W